Radial Basis Function Neural Networks Classification for the Recognition of Idiopathic Pulmonary Fibrosis in Microscopic Images

This study investigates the potential of applying the radial basis function (RBF) neural network architecture for the classification of biological microscopic images displaying lung tissue sections with idiopathic pulmonary fibrosis. For the development of the RBF classifiers, the fuzzy means clustering algorithm is utilized. This method is based on a fuzzy partition of the input space and requires only a short amount of time to select both the structure and the parameters of the RBF classifier. The new technique was applied in lung sections acquired using a microscope and captured by a digital camera, at a magnification of 4times. Age-and sex-matched, 6-to 8-week-old mice (five for each time point and five as control) were used for the induction of pulmonary fibrosis (cf. bleomycin). Bleomycin administration initially induces lung inflammation that is followed by a progressive destruction of the normal lung architecture. The captured images correspond to 7,15, and 23 days after bleomycin or saline injection and bronchoalveolar lavage (BAL) has been performed to the mice sample. The images were analyzed and color features were extracted. A support vector machines (SVMs)-based classifier was also employed for the same problem. The resulting scores derived by visual assessment of the images by expert pathologists were compared with the RBF and SVM classification outcome. Overall, the RBF neural network had a slightly better performance than that of the SVM classifier, but both performed very well, matching to a great percentage the scoring of the experts. There are some erroneous predictions of the algorithm for the regions characterized as "ill" regions (i.e., some bronchia were wrongly classified as fibrotic areas); however, in general, the algorithm worked pretty fine in distinguishing pathologic from normal in most cases and for heterogeneous fibrotic foci, achieving high values in terms of specificity and sensitivity.